Information processing method, device and equipment

Abstract

The present invention discloses an information processing method, apparatus, and device. The method comprises: obtaining serialized information of a first-type message; inputting the serialized information of the first-type message into a trained conditional adversarial network model for processing to obtain a second-type message; the conditional adversarial network model is trained based on a training sequence consisting of historical first-type messages and second-type messages obtained by authentically converting the historical first-type messages. This method can convert ordinary text short messages, which are first-type messages, into multimedia-rich second-type messages, improving the end-user experience.

Description

Technical Field

[0001] The present invention relates to the field of information processing technology, and in particular to an information processing method, apparatus and device. Background Art

[0002] 5G messaging, based on the native SMS interface of mobile devices, allows users to send and receive text, images, audio, video, location, contacts, and other media content. This includes point-to-point messaging, group messaging, group chat messaging, and messaging between peers and applications. Compared to the single-function traditional SMS, 5G messaging not only broadens the scope of information sending and receiving, supporting users with multimedia content such as text, audio and video, cards, and location, but also extends the depth of the interactive experience. Users can complete service search, discovery, interaction, payment, and other services within the message window, creating a one-stop service information window.

[0003] In existing technology, when the calling user is a regular text message sender and the called party is a 5G terminal, the 5GMC SMS processing module directly completes the subsequent SMS routing according to the existing IP SMS or circuit domain SMS process. However, the current conversion of regular text messages to 5G messages simply delivers the regular text message in text form to the called 5G terminal. This method fails to reflect the characteristics and advantages of 5G message rich media and does not provide a good user experience for 5G terminals. Summary of the Invention

[0004] In view of the above problems, embodiments of the present invention are proposed to provide an information processing method, apparatus, and device that overcome the above problems or at least partially solve the above problems.

[0005] According to one aspect of an embodiment of the present invention, there is provided an information processing method, including:

[0006] Get serialization information of the first type of message;

[0007] The serialized information of the first type of message is input into a trained conditional adversarial network model for processing to obtain a second type of message; the conditional adversarial network model is trained based on a training sequence consisting of historical first type messages and second type messages obtained after the historical first type messages are truly converted.

[0008] According to another aspect of an embodiment of the present invention, there is provided an information processing apparatus, including:

[0009] An acquisition module, configured to acquire serialization information of a first type of message;

[0010] A processing module is used to input the serialized information of the first type of message into a trained conditional adversarial network model for processing to obtain a second type of message; the conditional adversarial network model is trained based on a training sequence consisting of historical first type messages and second type messages obtained after the historical first type messages are truly converted.

[0011] According to another aspect of an embodiment of the present invention, there is provided a computing device, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus;

[0012] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned information processing method.

[0013] According to another aspect of the embodiments of the present invention, a computer storage medium is provided, wherein the storage medium stores at least one executable instruction, and the executable instruction enables a processor to perform operations corresponding to the above-mentioned information processing method.

[0014] According to the solution provided by the above embodiment of the present invention, the serialization information of the first type of message is obtained; the serialization information of the first type of message is input into the trained conditional adversarial network model for processing to obtain the second type of message; the conditional adversarial network model is trained based on a training sequence composed of historical first type messages and second type messages obtained after the real conversion of the historical first type of message. Ordinary text short messages as first type messages can be used as conditions to build a conditional generative adversarial network, and the trained generator therein can be used to convert the ordinary text short messages as first type messages into second type messages in a rich media format, such as 5G messages, and add pictures that match the text on the basis of the text short messages. It enriches the expression content required to carry the text and improves the experience of 5G terminal users.

[0015] The above description is only an overview of the technical solutions of the embodiments of the present invention. In order to more clearly understand the technical means of the embodiments of the present invention, they can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the embodiments of the present invention more obvious and easy to understand, the specific implementation methods of the embodiments of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the embodiments of the present invention. The same reference numerals are used throughout the accompanying drawings to denote the same components. In the accompanying drawings:

[0017] Figure 1 shows a flow chart of an information processing method provided by an embodiment of the present invention;

[0018] Figure 2 A schematic diagram of the architecture of a generator provided by an embodiment of the present invention is shown;

[0019] Figure 3 A flowchart of the conditional network adversarial model training provided by an embodiment of the present invention is shown;

[0020] Figure 4 A schematic diagram of the architecture of a conditional generative adversarial network model provided by an embodiment of the present invention is shown;

[0021] Figure 5 A flowchart of online conversion of ordinary text short messages to 5G messages provided by an embodiment of the present invention is shown;

[0022] Figure 6 A schematic structural diagram of an information processing device provided by an embodiment of the present invention is shown;

[0023] Figure 7 A schematic structural diagram of a computing device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0024] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0025] Figure 1 FIG. 1 is a flow chart showing an information processing method according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0026] Step 11, obtaining serialization information of the first type of message;

[0027] Specifically, the first type of message can be an ordinary text short message, and the first type of message is cleaned and serialized. For example, the first type of message is case-unified, uppercase letters are converted to lowercase letters, and each word in the first type of message is converted into an integer sequence, for example: ["qci": 40, "imsi": 105, "info": 8, "update": 278, "on": 89, "agent": 164, "modify": 59, "the": 21, "type": 303, "storage": 231, ...]. At the same time, the longest length K in the first type of message set is taken as its encoding sequence length, and the encoding sequence length of each first type of message is padded to K, to obtain the serialization information of the first type of message, which can be expressed as {c1 i , c2 i , c3 i ,…c k i}.

[0028] Step 12: Input the serialized information of the first type of message into the trained conditional adversarial network model for processing to obtain the second type of message; the conditional adversarial network model is trained based on a training sequence consisting of historical first type messages and second type messages obtained after the real conversion of the historical first type of messages.

[0029] The information processing method described in this embodiment obtains serialized information of a first type of message; the serialized information of the first type of message is input into a trained conditional adversarial network model for processing to obtain a second type of message. The caller-initiated text message can be used as a condition to build a conditional generative adversarial network, and the trained generator can be used to automatically generate a 5G message in a rich media format based on the text message, adding pictures that match the text and adding a suitable text presentation format. This solves the problem that converting ordinary text messages to 5G messages is simply delivering ordinary text messages to the called 5G terminal in the form of text, and enriches the expression content required to carry the text. The beneficial effect of increasing the user experience of the 5G terminal is achieved.

[0030] In one embodiment of the present invention, step 12 may include:

[0031] Step 121: Input the serialized information of the first type of message into the embedding layer of the generator of the conditional adversarial network model for processing to obtain a feature vector of the first type of message. The feature vector may be a vector of a fixed shape and 128 dimensions.

[0032] like Figure 2The figure shows the architecture of the generator of the conditional adversarial network model. The generator consists of an LSTM neural network with an encoder-decoder architecture. The neurons selected are long short-term memory (LSTM) neurons. LSTM can remember long-term information by controlling the storage time of the values ​​in the cache, which is suitable for learning long sequences. The generator includes 1 embedding layer, 12 hidden layers, and 1 fully connected layer. Among the 12 hidden layers, the first 6 layers belong to the encoder, and the last 6 layers belong to the decoder. The encoder contains two branches, which respectively input the encoded first type message and random noise z (dimension is 100, randomly drawn from a normal distribution), and the model parameters of each branch are set to the same. Then, the target second type message corresponding to the first type message initiated by the caller is generated through the LSTM neural network. Each of the 12 hidden layers is set with 64 LSTM neurons, and the activation function of each layer is set to "relu". The output layer (fully connected Dense layer) contains Dense fully connected neurons with the same size as the output sequence dimension, and the activation function is set to "softmax". The softmax output result is fed into the multi-class cross entropy loss function to convert the output shape into the final output dimension.

[0033] Specifically, the serialization information of the first type of message is input into the embedding layer of the generator of the conditional adversarial network model. The input data dimension is set to the text dictionary size of the first type of message set, and the output is set to the size of the vector space required to convert the word to 128 dimensions. The function of the embedding layer is to perform vector mapping (word embeddings) on each word in the input first type of message serialization, that is, to convert the integer sequence of each word in the first type of message serialization into a fixed-shape 128-dimensional vector.

[0034] Step 122: input the feature vector into the encoder of the generator of the conditional adversarial network model for processing to obtain a first output, which is Figure 2 The latent feature vector in ;

[0035] Specifically, the feature vector of the first type of message is input into the first branch of the encoder of the generator of the conditional adversarial network model for processing, and random noise is input into the second branch of the encoder of the generator of the conditional adversarial network model for processing, thereby obtaining a first output of the encoder; the first branch and the second branch each include N sequentially connected hidden layers, the neurons in the hidden layers are long short-term memory (LSTM) neurons, and N is a positive integer. Each hidden layer is provided with 64 LSTM neurons, and the activation function of each layer is set to "ReLU".

[0036] Step 123: Input the first output into the decoder of the generator of the conditional adversarial network model for processing to obtain a second type of message. Specifically, step 123 may include:

[0037] Step 1231: input the first output of the encoder into the decoder of the generator of the conditional adversarial network model for processing to obtain the second output of the decoder;

[0038] Specifically, the second output of the encoder is input into a decoder of the generator of the conditional adversarial network model for processing to obtain an output of the decoder. The decoder includes N sequentially connected hidden layers, each of which is a long short-term memory (LSTM) neuron, where N is a positive integer. Each hidden layer has 64 LSTM neurons, and the activation function of each layer is set to "ReLU."

[0039] Step 1232: Input the second output of the decoder into the fully connected layer of the generator of the conditional adversarial network model for processing to obtain the second type of message. The decoder includes N sequentially connected hidden layers, where the neurons in the hidden layers are long short-term memory (LSTM) neurons, and N is a positive integer.

[0040] Specifically, the second output of the decoder is input into the fully connected network of the generator of the conditional adversarial network model for processing to obtain the second type of message. The fully connected layer (Dense layer) of the decoder, also known as the output layer, contains a number of Dense fully connected neurons set to the same size as the output sequence dimension, and the activation function is set to "softmax". The softmax output result is fed into a multi-class cross entropy loss function to convert the output shape into the final output dimension.

[0041] Figure 3 FIG1 shows a flow chart of the conditional network adversarial model training provided by an embodiment of the present invention. Figure 3 As shown, the method includes the following steps:

[0042] Step 31: Acquire historical first-type messages and historical second-type messages for training, where the historical second-type messages for training are historical second-type messages obtained by authentically converting the historical first-type messages.

[0043] Step 32: inputting the serialized information of the historical first type messages for training into the generator of the conditional adversarial network model for training processing to obtain the target second type messages;

[0044] Step 33: Input the training sequence consisting of the historical first-type messages used for training as conditions and the target second-type messages together with the historical second-type messages used for training into the discriminator of the conditional adversarial network model for processing to obtain a discrimination result, and update the parameters of the discriminator according to the discrimination result;

[0045] Step 34: Update the parameters of the generator according to the parameters of the discriminator to obtain the trained conditional adversarial network model.

[0046] In yet another embodiment of the present invention, step 33 may include:

[0047] Step 331: Serialize the historical first-type messages used for training and the historical second-type messages used for training to obtain a training sequence;

[0048] Specifically, first collect the original text SMS messages sent by historical callers from the existing network 5GMC i (i.e. the first type of message mentioned above) and the corresponding real converted rich media 5G message x i (ie the second type of message) and marked as 1, collect the original text message c sent by the historical caller i (ie the first type of message mentioned above) and the converted rich media 5G message x' generated by the corresponding generator i (ie, the target second type message) and marked as 0.

[0049] Secondly, the data set is cleaned and the text is serialized. Unify the letter case, convert uppercase letters to lowercase letters, and convert each word in the data into an integer sequence, for example: ["qci": 40, "imsi": 105, "info": 8, "update": 278, "on": 89, "agent": 164, "modify": 59, "the": 21, "type": 303, "storage": 231, ...]. At the same time, take the longest length K in the caller-originated text message set (i.e., the first type of message set) as its encoding sequence length, and fill the length of each text with K, take the longest length M in the converted rich media 5G message set (i.e., the second type of message set) as its encoding sequence length, and fill the length of each text with M. Among them, the text message (first type of message) can be represented as {c1 i , c2 i , c3 i ,…c K i}; The converted rich media 5G message (second type message) contains the original text content and the added rich media format XML document matching the content, which can be represented as {x1i , x2 i , x3 i ,…x M i}, where the training sequence is a training sequence of length K+M.

[0050] Finally, the total dataset used for training is divided into a training set and a test set, with 90% of the total dataset being the training set and 10% being the test set. The training set is used to train the model, and the test set is used to test the model.

[0051] Step 332: input the training sequence and the target second type message into the embedding layer of the discriminator for processing to obtain a target feature vector;

[0052] Specifically, the serialized information of the text messages used for training (i.e., the first type of messages) and the serialized information of the converted rich media 5G messages (i.e., the second type of messages) are concatenated. The length of the concatenated sequence is K+M. The concatenated sequence is then input into the embedding layer of the generator of the conditional adversarial network model. The input data dimension is set to the total data text dictionary size, and the output is set to the size of the vector space required to convert the words to 128 dimensions. The caller-originated text message (i.e., the first type of message) text sequence is converted into a fixed-shape 128-dimensional vector.

[0053] Step 333: input the target feature vector into the convolution layer of the discriminator for processing to obtain the convolution layer output;

[0054] Specifically, the feature vector is input into the convolutional layer of the discriminator for text feature extraction to obtain the convolutional layer output. The number of convolution kernels in the convolution layer is 48 (i.e., the dimension of the output), the spatial window length of the convolution kernel is set to 2 (i.e., the convolution kernel reads two words continuously at a time), and the activation function is set to "ReLU".

[0055] Step 334: input the output of the convolutional layer into the maximum pooling layer of the discriminator for processing to obtain the maximum pooling layer output;

[0056] Specifically, the maximum value among the eigenvalues ​​extracted by the convolution kernel is retained, and all other eigenvalues ​​are discarded to obtain the output of the maximum pooling layer. The maximum pooling layer (MaxPooling1D) has a pooling window size of 2.

[0057] Step 335: The output of the maximum pooling layer is input into the flattening layer of the discriminator for processing to obtain a flattening layer output;

[0058] Specifically, the output of the max pooling layer is converted into two dimensions to obtain the output of the flattening layer. The flattening layer is used to "flatten" the output of the max pooling layer, converting the three-dimensional input into two dimensions, and is often used in the transition from the convolutional layer to the fully connected layer.

[0059] Step 336: The output of the flattening layer is input into the fully connected layer of the discriminator for processing to obtain the output of the fully connected layer;

[0060] Specifically, the fully connected layer includes 16 neurons, and the activation function is set to "relu".

[0061] In step 337 , the output of the fully connected layer is input into the objective function for processing to obtain a discrimination result.

[0062] The objective function is:

[0063] Among them, θ is the model parameter, c i is the i-th first type message, x i For the i-th second type message, D(c i , x i ) is the output value of the discriminator, when D(c i , x i ) is positive, the discriminator parameters are updated to increase P θ (x i |c i ), when D(c i , x i ) is negative, the discriminator parameters are updated to reduce P θ (x i |c i ).

[0064] Specifically, the output layer (Dense fully connected layer) contains one Dense neuron, the activation function is set to "sigmoid", and the output value is 1 or 0. The sigmoid output result is input into the binary cross entropy loss function to obtain the discrimination result.

[0065] like Figure 4The following figure shows the architecture of a conditional network adversarial model. The parameters of the generator are fixed, and the discriminator is trained. The original text message c (i.e., the first type message) sent by the caller and the corresponding real converted enriched 5G message x (i.e., the second type message) are randomly extracted from the training data. The original text message c (i.e., the first type message) sent by the caller is input into the generator G(c), which generates the corresponding converted enriched 5G message x' (i.e., the target second type message). The original text message c (i.e., the first type message) sent by the caller is used as a condition and input into the discriminator D(c, x') along with the corresponding generated converted enriched 5G message x' (i.e., the target second type message). The discriminator learns to give a higher score to the pair of the original text message sent by the caller (i.e., the first type message) and the corresponding real converted enriched 5G message (i.e., the second type message), and a lower score to the pair of the original text message sent by the caller (i.e., the first type message) and the corresponding generated converted enriched 5G message (i.e., the target second type message), thereby updating the discriminator parameters.

[0066] The parameters of the discriminator are fixed, and the generator is trained so that the converted rich media 5G message (i.e., the target second type message) generated by the generator is increasingly similar to the real converted rich media 5G message (i.e., the second type message), thereby maximizing the score output by the discriminator (true is 1, false is 0). The score can be regarded as a reward. The parameters of the generator are updated according to the score output by the maximized discriminator to obtain a trained conditional adversarial network model.

[0067] In this embodiment, the adversarial network model (GAN, Generative adversarial network) mainly includes two parts: a generator (G) and a discriminator (D). The two are in a competitive adversarial relationship. First, a discriminator that can distinguish between real data and generated data is added to the generation process, so that the generator G and the discriminator D compete with each other. Among them, the role of the discriminator D is to distinguish between real data and generated data, and the role of the generator G is to improve itself so as to generate data that can deceive the discriminator D. Secondly, the generator and the discriminator are constantly in competition. When the discriminator D can no longer distinguish between true and false data, the trained generator G is obtained. The conditional adversarial network model (Conditional GAN) adds a condition y to the adversarial network model (GAN, Generative adversarial network). The condition is added to the generator and discriminator for training, so that the trained model can generate the required data according to the set conditions. Both the generator and the discriminator add additional information y as a condition. y can be any information, such as category information, or data of other modalities. The training of the conditional adversarial network model is achieved by feeding additional information y to the discriminative model and the generative model as part of the input layer.

[0068] like Figure 5 As shown, in an embodiment of the present invention, a specific implementation process of an information processing method is provided. The first type of message takes an ordinary text short message as an example, and the second type of message takes a 5G message as an example, including the following:

[0069] Step 51: The calling user (a 5G terminal residing in the circuit domain, a volte terminal residing in the circuit domain, or a 2 / 3G terminal) sends an SMS-submit message to the VMSC to initiate the SMS service process.

[0070] Step 52: The VMSC sends the SMS message to the SMS processing function in the calling 5GMC via MAP-MO-forward-short-message.

[0071] Step 53: The 5GMC SMS processing function sends the SMS to the SMS pre-processing module to serialize the text.

[0072] Step 54: Send the serialized text message and random noise to the text message to rich media 5G message generator composed of the pre-trained generator G;

[0073] Step 55: The text message to rich media 5G message converter uses a codec neural network composed of long short-term memory neurons to automatically generate a rich media 5G message based on the text message and random noise, adding images that match the text and adding a suitable text presentation format (such as a card) to more richly carry the content required by the text;

[0074] Step 56: Send the generated rich media format 5G message to the message processing function of the calling 5GMC;

[0075] Step 57: Finally, the rich media format 5G message is sent to the called 5G terminal through the access function of 5GMC.

[0076] In the above embodiment of the present invention, a conditional generative adversarial network (conditional GAN) is constructed by taking the caller-originated text message as a condition, and using the trained generator to automatically generate a 5G message in a rich media format based on the text message and random noise, adding a presentation format that matches the text (such as a card, etc.) to more richly carry the content required to express the text. The conditional generative adversarial network model designed in this proposal consists of a generator and a discriminator, wherein the text message to rich media 5G message generator G is composed of an LSTM neural network with an encoder-decoder architecture, responsible for generating a rich media 5G message corresponding to the caller-originated text message; wherein the discriminator D is composed of a convolutional neural network, responsible for judging the difference between the generated result and the actual converted rich media 5G message, and adjusting the generator G according to the score output by the discriminator. The generator is then adjusted according to the output of the discriminator so that it can generate a correct rich media 5G message corresponding to the caller-originated text message. This improves the 5G message experience of 5G terminal users.

[0077] Figure 6 FIG. 6 is a schematic diagram showing the structure of an information processing device 60 provided by an embodiment of the present invention. Figure 6 As shown, the device includes:

[0078] An acquisition module 61 is configured to acquire serialization information of a first type of message;

[0079] The processing module 62 is used to input the serialized information of the first type of message into the trained conditional adversarial network model for processing to obtain the second type of message; the conditional adversarial network model is trained based on a training sequence consisting of historical first type messages and second type messages obtained after the actual conversion of the historical first type of messages.

[0080] Optionally, the processing module 62 is further configured to input the serialized information of the first type of message into the embedding layer of the generator of the conditional adversarial network model for processing to obtain a feature vector of the first type of message;

[0081] Inputting the feature vector into an encoder of a generator of the conditional adversarial network model for processing to obtain a first output;

[0082] The first output is input into the decoder of the generator of the conditional adversarial network model for processing to obtain a second type of message.

[0083] Optionally, the processing module 62 is also used to input the feature vector of the first type of message into the first branch of the encoder of the generator of the conditional adversarial network model for processing, and to input random noise into the second branch of the encoder of the generator of the conditional adversarial network model for processing to obtain the first output of the encoder; the first branch and the second branch both include N sequentially connected hidden layers, the neurons of the hidden layers are long short-term memory (LSTM) neurons, and N is a positive integer.

[0084] Optionally, the processing module 62 is further configured to input the first output of the encoder into a decoder of the generator of the conditional adversarial network model for processing to obtain a second output of the decoder;

[0085] The second output of the decoder is input into the fully connected layer of the generator of the conditional adversarial network model for processing to obtain the second type message; the decoder includes N sequentially connected hidden layers, the neurons of the hidden layers are long short-term memory (LSTM) neurons, and N is a positive integer.

[0086] Optionally, the processing module 62 is further configured to obtain historical first-type messages for training and historical second-type messages for training, wherein the historical second-type messages for training are historical second-type messages obtained by authentically converting the historical first-type messages;

[0087] Inputting the serialized information of the historical first type messages for training into the generator of the conditional adversarial network model for training processing to obtain the target second type messages;

[0088] Inputting the training sequence composed of the historical first-type messages used for training as conditions and the target second-type messages together into the discriminator of the conditional adversarial network model for processing to obtain a discrimination result, and updating the parameters of the discriminator according to the discrimination result;

[0089] According to the parameters of the discriminator, the parameters of the generator are updated to obtain the trained conditional adversarial network model.

[0090] Optionally, the processing module 62 is further configured to serialize the historical first-type messages used for training and the historical second-type messages used for training to obtain a training sequence;

[0091] Inputting the training sequence and the target second type message into the embedding layer of the discriminator for processing to obtain a target feature vector;

[0092] Input the target feature vector into the convolution layer of the discriminator for processing to obtain the convolution layer output;

[0093] The output of the convolutional layer is input into the maximum pooling layer of the discriminator for processing to obtain the maximum pooling layer output;

[0094] The output of the maximum pooling layer is input into the flattening layer of the discriminator for processing to obtain a flattening layer output;

[0095] The output of the flattening layer is input into the fully connected layer of the discriminator for processing to obtain the output of the fully connected layer;

[0096] The output of the fully connected layer is input into an objective function for processing to obtain a discrimination result.

[0097] Optionally, the objective function is:

[0098]

[0099] Among them, θ is the model parameter, c i is the i-th historical first type message, x i For the i-th historical second type message, D(c i , x i ) is the output value of the discriminator, when D(c i , x i ) is positive, the discriminator parameters are updated to increase P θ (x i |c i ), when D(c i , x i ) is negative, the discriminator parameters are updated to reduce P θ (x i |c i ).

[0100] It should be noted that this embodiment is an apparatus embodiment corresponding to the above method embodiment, and all implementation methods in the above method embodiment are applicable to the embodiment of this apparatus and can achieve the same technical effects.

[0101] An embodiment of the present invention provides a non-volatile computer storage medium, wherein the computer storage medium stores at least one executable instruction, and the computer executable instruction can execute the information processing method in any of the above method embodiments.

[0102] Figure 7 The schematic diagram of the structure of the computing device provided by the embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computing device.

[0103] like Figure 7 As shown, the computing device may include: a processor, a communication interface, a memory, and a communication bus.

[0104] The processor, communication interface, and memory communicate with each other via a communication bus. The communication interface is used to communicate with other devices, such as clients or other server network elements. The processor is used to execute programs, specifically, the steps described in the aforementioned information processing method embodiment for a computing device.

[0105] Specifically, the program may include program codes including computer operation instructions.

[0106] The processor may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the computing device may be processors of the same type, such as one or more CPUs, or processors of different types, such as one or more CPUs and one or more ASICs.

[0107] The memory is used to store programs. The memory may include high-speed RAM memory and may also include non-volatile memory (non-volatile memory), such as at least one disk storage.

[0108] The program can be specifically used to cause the processor to execute the information processing method in any of the above-mentioned method embodiments. The specific implementation of each step in the program can refer to the corresponding description of the corresponding steps and units in the above-mentioned information processing method embodiments, and will not be repeated here. Those skilled in the art will clearly understand that for the convenience and brevity of description, the specific working process of the above-mentioned devices and modules can refer to the corresponding process description in the above-mentioned method embodiments, and will not be repeated here.

[0109] The algorithm or display provided herein is not inherently related to any particular computer, virtual system or other device. Various general-purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the embodiment of the present invention is not directed to any specific programming language. It should be understood that various programming languages ​​can be utilized to implement the content of the embodiment of the present invention described herein, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the embodiment of the present invention.

[0110] In the description provided herein, numerous specific details are described. However, it is understood that embodiments of the present invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of this description.

[0111] Similarly, it should be understood that in order to streamline the embodiments of the invention and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the invention, various features of the embodiments of the invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed approach should not be interpreted as reflecting an intention that the claimed embodiments of the invention require more features than are expressly recited in each claim. Rather, as reflected in the claims below, inventive aspects lie in less than all of the features of the individual embodiments disclosed above. Accordingly, the claims that follow the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of the invention.

[0112] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition may be divided into multiple submodules or subunits or subcomponents. All features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed herein may be combined in any combination, except that at least some of such features and / or processes or units are mutually exclusive. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0113] Furthermore, those skilled in the art will appreciate that although some embodiments herein include certain features included in other embodiments but not other features, combinations of features from different embodiments are intended to be within the scope of the present invention and to form different embodiments. For example, in the claims below, any of the claimed embodiments may be used in any combination.

[0114] The various component embodiments of the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) can be used in practice to implement some or all of the functions of some or all of the components according to the embodiments of the present invention. The embodiments of the present invention can also be implemented as a device or apparatus program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program implementing an embodiment of the present invention can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0115] It should be noted that the above embodiments illustrate rather than limit the embodiments of the invention, and that a person skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between brackets should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The embodiments of the invention may be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third etc. does not indicate any order. These words may be interpreted as names. The steps in the above embodiments should not be understood as limiting the order of execution unless otherwise specified.

Claims

1. An information processing method, characterized in that: include: Get serialization information of the first type of message; Inputting serialized information of the first type of message into a trained conditional adversarial network model for processing to obtain a second type of message; the conditional adversarial network model is trained based on a training sequence consisting of historical first type messages and second type messages obtained by actual conversion of the historical first type of messages; The conditional adversarial network model is trained through the following process: Acquire historical first-type messages for training and historical second-type messages for training, where the historical second-type messages for training are historical second-type messages obtained by authentically converting the historical first-type messages; Inputting the serialized information of the historical first type messages for training into the generator of the conditional adversarial network model for training processing to obtain the target second type messages; Inputting the training sequence composed of the historical first-type messages used for training as conditions and the target second-type messages together into the discriminator of the conditional adversarial network model for processing to obtain a discrimination result, and updating the parameters of the discriminator according to the discrimination result; According to the parameters of the discriminator, the parameters of the generator are updated to obtain the trained conditional adversarial network model; Among them, the discriminator includes in sequence: an embedding layer, a convolutional layer, a maximum pooling layer, a flattening layer and a fully connected layer; the embedding layer processes the input training sequence and the target second type message, and the output of the fully connected layer is input into the objective function for processing to obtain the discrimination result.

2. The information processing method according to claim 1, wherein: Inputting the serialized information of the first type of message into the trained conditional adversarial network model for processing to obtain the second type of message, including: Inputting the serialized information of the first type of message into the embedding layer of the generator of the conditional adversarial network model for processing to obtain a feature vector of the first type of message; Inputting the feature vector into an encoder of a generator of the conditional adversarial network model for processing to obtain a first output; The first output is input into the decoder of the generator of the conditional adversarial network model for processing to obtain a second type of message.

3. The information processing method according to claim 2, wherein: Inputting the feature vector into an encoder of a generator of the conditional adversarial network model for processing to obtain a first output includes: The feature vector of the first type of message is input into the first branch of the encoder of the generator of the conditional adversarial network model for processing, and random noise is input into the second branch of the encoder of the generator of the conditional adversarial network model for processing to obtain the first output of the encoder; the first branch and the second branch both include N sequentially connected hidden layers, the neurons of the hidden layers are long short-term memory (LSTM) neurons, and N is a positive integer.

4. The information processing method according to claim 2, wherein: Inputting the first output into a decoder of the generator of the conditional adversarial network model for processing to obtain a second type of message, including: Inputting the first output of the encoder into the decoder of the generator of the conditional adversarial network model for processing to obtain the second output of the decoder; The second output of the decoder is input into the fully connected layer of the generator of the conditional adversarial network model for processing to obtain the second type message; the decoder includes N sequentially connected hidden layers, the neurons of the hidden layers are long short-term memory (LSTM) neurons, and N is a positive integer.

5. The information processing method according to claim 1, wherein: The training sequence consisting of the historical first-type messages used for training as a condition and the historical second-type messages used for training and the target second-type message are input into the discriminator of the conditional adversarial network model for processing to obtain a discrimination result, including: Serializing the historical messages of the first type used for training and the historical messages of the second type used for training to obtain a training sequence; Inputting the training sequence and the target second type message into the embedding layer of the discriminator for processing to obtain a target feature vector; Input the target feature vector into the convolution layer of the discriminator for processing to obtain the convolution layer output; The output of the convolutional layer is input into the maximum pooling layer of the discriminator for processing to obtain the maximum pooling layer output; The output of the maximum pooling layer is input into the flattening layer of the discriminator for processing to obtain a flattening layer output; The output of the flattening layer is input into the fully connected layer of the discriminator for processing to obtain the output of the fully connected layer; The output of the fully connected layer is input into an objective function for processing to obtain a discrimination result.

6. The information processing method according to claim 5, characterized in that The objective function is: Among them, θ is the model parameter, c i is the i-th historical first type message, x i For the i-th historical second type message, D(c i , x i ) is the output value of the discriminator, when D(c i , x i ) is positive, the discriminator parameters are updated to increase P θ (x i |c i ), when D(c i , x i ) is negative, the discriminator parameters are updated to reduce P θ (x i |c i ).

7. An information processing device, characterized in that The device comprises: An acquisition module, configured to acquire serialization information of a first type of message; a processing module, configured to input serialized information of the first type of message into a trained conditional adversarial network model for processing to obtain a second type of message; the conditional adversarial network model is trained based on a training sequence consisting of historical first type messages and second type messages obtained by authentic conversion of the historical first type of messages; The processing module is further configured to: Acquire historical first-type messages for training and historical second-type messages for training, where the historical second-type messages for training are historical second-type messages obtained by authentically converting the historical first-type messages; Inputting the serialized information of the historical first type messages for training into the generator of the conditional adversarial network model for training processing to obtain the target second type messages; Inputting the training sequence composed of the historical first-type messages used for training as conditions and the target second-type messages together into the discriminator of the conditional adversarial network model for processing to obtain a discrimination result, and updating the parameters of the discriminator according to the discrimination result; According to the parameters of the discriminator, the parameters of the generator are updated to obtain the trained conditional adversarial network model; Among them, the discriminator includes in sequence: an embedding layer, a convolutional layer, a maximum pooling layer, a flattening layer and a fully connected layer; the embedding layer processes the input training sequence and the target second type message, and the output of the fully connected layer is input into the objective function for processing to obtain the discrimination result.

8. A computing device comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute an operation corresponding to the information processing method according to any one of claims 1 to 6.

9. A computer storage medium, wherein at least one executable instruction is stored in the storage medium, and wherein the executable instruction enables a processor to execute an operation corresponding to the information processing method according to any one of claims 1 to 6.

Images